MATERIALS AND METHODS
The data used in this study was obtained from a experiment carried out at Moorepark Research Centre in the Republic of Ireland over a 5-yr period (2001-2005) comparing three strains of Holstein-Friesian cows in three feeding systems (Horan et al. 2005; McCarthy et al., 2007)
Animals
The three strains of Holstein-Friesian (HF) cows were compared: high production North American (HP), high durability North American (HD) and New Zealand (NZ). The HP strain represents the outcome of continuous aggressive selection for milk production; the HD strain represents a more balanced breeding policy, which also considers other traits like fertility when selecting milk production; the NZ strain represents the high NZ genetic cows (Horan et al. 2005; McCarthy et al., 2007).
To create the HP strain, the top 50% of HF cows in the Moorepark herds (based on pedigree index for milk production) were inseminated with semen from five North American Holstein-Friesian sires. These five sires were selected on predicted transmitting ability for milk, fat, and protein yields (Horan et al. 2005; McCarthy et al., 2007).
To create the HD strain, the bottom 50% of HF cows in the Moorepark herds (based on pedigree index for milk production) were inseminated with semen from five North American Holstein-Friesian sires. These five sires were selected base on combined pedigree indices for milk production, fertility and linear traits (Horan et al. 2005; McCarthy et al., 2007).
The NZ cows originated from embryos imported from NZ. The embryos were created by mating high genetic merit NZHF cows to five high genetic merit NZHF sires (Horan et al. 2005; McCarthy et al., 2007).
27 Feeding systems
The three feeding systems were: a high grass allowance feeding system which is typical of spring calving herds in Ireland (MP, control); a higher concentrate system (HC) and a higher stocking rate system (HS) (Horan et al. 2005; McCarthy et al., 2007). In the control group (MP), the overall stocking rate (SR) was 2.47 cows/ha with a nitrogen (N) fertilizer input of 290 kg N/ha (from early January to late September) and received 368 kg concentrate per cow during early lactation, the remainder of diet still came from grazed pasture. The HC group had similar SR and N input to the control group but with a higher concentrate input of 1,452 kg/cow. The HS group had similar concentrate (327 kg/cow) and N input to the control group but at a higher SR of 2.74 cows/ha (Horan et al. 2005; McCarthy et al., 2007).
The MP and HC feeding systems were designed to allow each strain to express its potential within each feed system largely unrestricted by limitations in feed supply. The HS feeding system was aim to reduce feed allowance by increasing SR (Horan et al. 2005; McCarthy et al., 2007).
A total of 99, 117, 117, 124 and 126 animals were used in Year 1, 2, 3, 4 and 5, respectively, divided between strains and feeding systems (Table 3.1).
Table 3.1. The number of dairy cow records included in the strain/feeding system trial.
Strain of Holstein-Friesian1 Feed system2
Group HP HD NZ MP HS HC No. of animals Year 1 33 33 33 33 33 33 Year 2 39 39 39 39 39 39 Year 3 39 39 39 39 39 39 Year 4 40 42 42 40 42 42 Year 5 36 48 42 42 42 42 1
HP = High production; HD = high durability; NZ = New Zealand. 2
MP = High grass feed system; HS = high stocking rate feed system; HC = high concentration feed system.
28 Animal measurements
Cows were milked twice daily in a 14-unit highline milking parlour. Milk yield was measured at each milking for each cow using electronic milk meters (McCarthy et al., 2007). The concentrations of fat, protein, and lactose in milk were determined using a Milkoscan 203 (Foss Electric DK-3400, Hillerød, Denmark) from successive morning and evening samples collected once weekly (Horan et al., 2005). The BCS was recorded every 3 weeks during lactation on a scale of 0 to 5 (0 = emaciated, 5 = extremely fat) with increments of 0.25. Fertility parameters included the interval from calving to first service, the interval from calving to conception, number of services per conception, pregnant to first service, pregnant to second service, pregnant to first and second services, submission within 3 weeks of the start of the breeding season and overall pregnancy rate.
Statistical Analysis
Lactation curves for daily yields of milk, fat, protein and lactose and BCS were modelled using an orthogonal polynomial of order 2:
e p p
yt =α0+α1 t+α2 t2+
where y is the daily yield of milk, fat, protein or lactose at day t. Legendre polynomials
are conventionally defined in the range -1 ≤t≤ +1 and are orthogonal within this
range; days in milk were standardized to lie between -1 and +1 before evaluating
the polynomials. Functions of pt of Legendre polynomials were calculated as P0 (t)
= 1, P1 (t) = xt, P2 (t) =1/2(3xt2-1), with xt = (2*t - (305 + 1)) / (305 - 1), and α0, α1
and α2 are regression coefficients.
Estimates of α0, α1 and α2 for each lactation of each trait were obtained using PROC
GLM of SAS based on the herd test records (SAS Institute, 2002). With these α0, α1 and
α2 values, the lactation curves for milk, fat, protein and lactose and BCS were modelled.
Total lactation yields of milk, fat, protein and lactose were obtained as the sum of the yield of each lactating day. Accumulated yields were calculated at different periods postpartum: first 60, first 120, first 180 and actual lactation length. Based on these yields, protein percentages and fat to protein ratio during different periods postpartum were calculated. Different measures of BCS (BCS at calving, BCS on Day 60 and BCS
29
change from calving to Day 60) for each cow on each day of lactation were obtained using the lactation curve for BCS.
Pearson correlations among production, fertility and BCS traits were estimated. Logistic regression analysis (Hosmer and Lemeshow, 1989) was carried out in PROC GENOMD (SAS Institute, 2002) to investigate the associations between milk protein percentage and the binary fertility traits. The logistic regression model considered the fixed effect of genotype, feeding system, the interaction between genotype and feeding system, lactation number and year and the random effect of year.
The prediction of probability of reproductive performance was using the following model: ) ... ( - 0 11 2 2 k e 1 1 k x x x P β +β +β β + =
where P is the probability of reproductive performance, in the range 0 ≤P≤ +1.β0 is a
constant and βk are coefficients of the predictor variables.
Another logistic regression analysis (Hosmer and Lemeshow, 1989) was carried out in PROC GENOMD (SAS Institute, 2002) to investigate the associations between BCS and production traits considered as independent variables and the binary fertility traits considered as dependent variables following a binomial distribution. The prediction of probability of reproductive performance was using the same model as shown above. Because of a large correlation between BCS traits and production traits, principal components analysis was applied to represent the BCS and production traits.
Principal component analysis is a method of transforming the original independent variables into new, uncorrelated variables (Lafi and Kaneene, 1992). The new, uncorrelated variables are called principal components. Each principal component is a linear combination of all the original independent variables.
The purpose of using principal component analysis is to reduce the dimensionality without losing much of the information. Since principal components labelled according to the size of their variance, the first principal component explains the largest amount of variation among the variables while the last principal component explains the least. Normally, the first few principal components are the most informative ones. Therefore,
30
the reduction of dimensionality can be achieved by choosing only the first few principal components to be used in the regression analysis (Lafi and Kaneene, 1992).
Let X be the original data set, and Y is a re-representation of that data set, which is
related by a linear transformation P.
PX = Y
If pi are the rows of P, xi are the columns of X (or individual 𝑋𝑋⃗ ), and yi are the
columns of Y, then 𝑃𝑃𝑋𝑋 = �𝑝𝑝⋮1 𝑝𝑝𝑚𝑚 � [𝑥𝑥1 … 𝑥𝑥𝑛𝑛] 𝑌𝑌 = �𝑝𝑝1∙ 𝑥𝑥⋮ 1 ⋯ 𝑝𝑝⋱ 1∙ 𝑥𝑥⋮ 𝑛𝑛 𝑝𝑝𝑚𝑚 ∙ 𝑥𝑥1 ⋯ 𝑝𝑝𝑚𝑚 ∙ 𝑥𝑥𝑛𝑛 � 𝑦𝑦𝑖𝑖 = � 𝑝𝑝1∙ 𝑥𝑥𝑖𝑖 ⋮ 𝑝𝑝𝑚𝑚 ∙ 𝑥𝑥𝑖𝑖 �
In present study, principal components y1, y2, ..., ynwhich are uncorrelated with each
other and account for decreasing proportions of the total variance of the original variables were defined as:
y1 = p11x1 + p12x2 + ... + p1nxn
y2 = p21x1 + p22x2 + ... + p2nxn
……
yn = pn1x1 + pn2x2 + ... + pnnxn
with coefficients p was the respective eigenvalues, so that y1, y2, ..., yn account for
decreasing proportions of the total variance of the original variables, x1, x2, ...,
xn.
Principal component for BCS traits was the re-representation of BCS at calving and BCS change; principal component for milk production was the re-representation of accumulated milk, fat, protein and lactose yield; principal component for protein percentage was the re-representation of milk protein percentage. For each principal
31
Institute, 2002). Fertility performance was then analysed with a logistic regression model that considered the fixed effect of year and lactation number as class effects and the principal component for milk production, principal component for BCS and principal component for protein percentage as covariables and the random effect of cow to account for repeated measures on the same cow.
Lactations were divided in the two groups according to protein percentage over the whole lactation, low or high. The classification of lactation into each of these groups was based on the average protein percentage for each combination of strain and feeding system. Overall rate was then analysed with a logistic regression model using the GENMOD procedure of SAS (2002). The model included the fixed effect of P% group, strain, feeding system, the interaction of strain and feeding system, lactation number and year.
32
Chapter4
RESULTS
DESCRIPTIVE STATISTICS
Descriptive statistics of milk production and fertility traits for each of the strains and feeding systems are given in Table 4.1. Amongst the different strains of HF cows, high production (HP) cows had the highest milk yield with New Zealand (NZ) cows had the lowest. However, the milk fat and protein percentage was greatest in NZ cows. High durability (HD) cows had the highest 21-day submission rate, while NZ cows had the highest first service and overall pregnant rates, as well as fewer services to achieve a success pregnancy.
Cows on the high concentrate (HC) feeding system produced more milk than cows on the other two feeding systems. As for reproductive performance, cows in the HC feeding system showed relatively high 21-day submission and first service conception rates, but similar overall pregnancy rates were found in both HC and Moorepark (MP) feeding system. Furthermore, cows in the MP feeding system needed the least number of services per pregnancy. Cows on the high stocking rate (HS) feeding system seemed to showed the worst reproductive performance.
33
Table 4.1. Mean (± SD) productive and fertility traits of three strains of Holstein-Friesians cows under three feeding systems.
HP1 HD2 NZ3 Effect MP4 HC5 HS6 MP HC HS MP HC HS Number of lactations 61 63 63 67 67 67 65 65 65 Milk production Milk, kg/cow 6346(1135) 7487(1135) 6420(1006) 6307(1024) 7202(1114) 6188(1019) 5881(1034) 6306(1165) 5601(906) Fat, g/kg 41.1(4.3) 39.0(3.4) 39.6(4.3) 39.5(2.7) 39.1(3.8) 40.8(3.9) 44.0(3.0) 44.2(3.7) 44.2(3.3) Protein, g/kg 34.6(1.7) 34.4(1.8) 34.0(1.8) 34.4(1.4) 35.2(2.0) 34.7(1.9) 35.9(1.6) 36.1(1.7) 35.5(2.2) Lactose, g/kg 46.7(1.0) 47.6(0.9) 46.8(1.2) 46.9(1.0) 47.2(1.2) 46.9(1.3) 47.2(1.1) 47.8(1.0) 47.0(1.2) Reproduction
21-day submission rate, % 77(42) 78(42) 74(44) 94(24) 91(29) 77(41) 86(35) 91(29) 84(36)
1st service conception rate, % 46(50) 49(50) 46(50) 58(49) 51(50) 48(50) 57(50) 70(46) 53(50)
Overall pregnancy rate, % 70(46) 76(43) 78(42) 95(21) 86(34) 74(44) 91(29) 94(24) 89(31)
Total services per cow 2.0(1.2) 2.0(1.1) 1.8(1.0) 1.6(0.9) 1.8(1.0) 1.9(1.1) 1.7(0.9) 1.5(1.0) 1.7(0.9)
1
HP = high production; 2HD = high durability; 3 NZ = New Zealand;
4
34
ASSOCIATIONS AMONG TRAITS
The relationship between BCS and milk production
Figure 4.1. Relationship between milk yield and body condition score (BCS) at calving for three different genotype cows under the same feeding system and same lactation. Geno1 represented the high durability North American Holstein-Friesian cows; Geno2 represented the high production North American Holstein-Friesian cows; Geno3 represented the New Zealand Holstein-Friesian cows.
The association between BCS at calving and milk production was non-linear. Milk production was positively associated with BCS at calving up to a BCS of about 2.5 to 2.75 units after which the association was negative. Similar trends were observed for BCS on Day 60. 4000 4500 5000 5500 6000 6500 7000 1.5 2 2.5 3 3.5 4 4.5
Geno1 Geno2 Geno3
BCS M ilk Y ie ld ( k g )
35
The relationship between BCS and reproductive performance
Table 4.2 shows estimates of Pearson correlations coefficients between measures of BCS and fertility traits. There were some significant correlations between measures of BCS and fertility traits.
Increased BCS at calving was associated with a reduced number of services (P<0.05), increased pregnancy rate to second service (P<0.05), increased pregnancy rate to first and second services (P<0.01), increased submission rate within 3 weeks after the start of breeding season (P<0.001), and increased overall pregnancy rate (P<0.001).
An increase of BCS on Day 60 was associated with an increased interval between calving and first service (P<0.01), increased pregnancy rate to first service (P<0.05), pregnancy rate to second service (P<0.05), pregnancy rate to first and second services (P<0.001), submission rate within 3 weeks after the start of breeding season (P<0.001), and overall pregnancy rate (P<0.001). Less number of service (P<0.05) were required for achieving a successful pregnancy as BCS_60 increased.
The greater the loss in BCS in early lactation, the longer the interval between calving and first service (P < 0.001) and between calving and conception (P < 0.001). There were no other significant correlations.
36
Table 4.2. Correlation coefficients between measures of body condition score and measures of fertility traits.
Trait BCS_1 BCS_60 BCS_CH CSI CCI NO_SERV PREG1 PREG2 PREG1_2 SERV_3wk PREG
BCS_11 1 BCS_602 0.82*** 1 BCS_CH3 -0.56*** 0.01 1 CSI4 0.00 0.12** 0.19*** 1 CCI5 -0.03 0.08 0.17*** 0.62*** 1 NO_SERV6 -0.10* -0.09* 0.04 -0.20*** 0.58*** 1 PREG17 0.08 0.09* -0.00 0.19*** -0.48*** -0.81*** 1 PREG28 0.12* 0.14* -0.01 0.10 -0.37*** -0.70*** . 1 PREG1_29 0.13** 0.14*** -0.02 0.18*** -0.48*** -0.82*** 0.63*** 0.98*** 1 SERV_3wk10 0.15*** 0.22*** 0.07 0.06 0.10* 0.06 0.04 0.00 0.03 1 PREG11 0.14*** 0.19*** 0.03 0.06 -0.19*** -0.46*** 0.44*** 0.41*** 0.52*** 0.18*** 1 * P<0.05; **P<0.01; ***P<0.001. 1
BCS_1 = body condition score at calving; 2BCS_60 = body condition score on Day 60;
3
BCS_CH = body condition score change (BCS_60 – BCS_1); 4CSI = calving to submission interval; 5CCI = calving to conception interval;
6
NO_SERV = number of services per conception; 7PREG1 = pregnant to first service; 8PREG2 = pregnant to second service;
9
37
Relationship between milk production and reproductive performance
Estimates of correlation coefficients between production and fertility traits are shown
in Table 4.3. Total milk, fat, protein and lactose yields were not correlated with calving to submission interval. However, milk, protein and lactose yields were positively correlated with calving to conception interval (P<0.05). Significant correlations were observed between yields in early lactation and calving to submission interval or calving to conception interval. The greater the yield of milk, fat, protein and lactose in the first 60 days of lactation, the shorter the calving to submission interval or calving to conception interval (P<0.001). Number of services per conception was positively correlated with milk (P<0.01) and lactose yields (P<0.05). Total milk, fat, protein and lactose yields were not correlated with first service pregnancy rate, but were negatively correlated with second service pregnancy rate (P<0.05). Negative correlations existed between both first and second service pregnancy rate and total milk (P<0.01), protein (P<0.05), fat (P<0.01) and lactose (P<0.05) yields, first 60 days milk (P<0.05), protein (P<0.05) and fat (P<0.05) yields, Overall pregnancy rate was negatively correlated with first 60 days milk (P<0.01), protein (P<0.01), fat (P<0.01) and lactose (P<0.01) yields, total milk (P<0.01), protein (P<0.05), fat (P<0.05) and lactose (P<0.01) yields. Reduced first three weeks submission rate was associated with increased first 60 days milk, protein, fat and lactose yields (P<0.001).
38
Table 4.3. Correlation coefficients between production traits and measures of fertility.
Trait MY MY_60 PY PY_60 FY FY_60 LY LY_60 P% P%_60 RATIO_FP CSI CCI NO_SERV PREG1 PREG2 PREG1_2 SERV_3wk PREG
MY1 1 MY_602 0.82*** 1 PY3 0.95*** 0.78*** 1 PY_604 0.77*** 0.97*** 0.79*** 1 FY5 0.84*** 0.76*** 0.89*** 0.76*** 1 FY_606 0.60*** 0.86*** 0.62*** 0.86*** 0.78*** 1 LY7 0.99*** 0.78*** 0.95*** 0.73*** 0.84*** 0.57*** 1 LY_608 0.82*** 0.99*** 0.79*** 0.96*** 0.76*** 0.85*** 0.79*** 1 P%9 -0.24*** -0.18*** 0.05 0.01 0.08* -0.00 -0.23*** -0.17*** 1 P%_6010 -0.10* 0.01 0.13** 0.26*** 0.11** 0.14*** -0.09* 0.02 0.75*** 1 RATIO_FP11 -0.29*** -0.09* -0.28*** -0.10* 0.19*** 0.33*** -0.28*** -0.08 -0.04 0.06 1 CSI12 0.04 -0.21*** 0.08 -0.23*** 0.06 -0.22*** 0.05 -0.21*** 0.11** -0.13** -0.06 1 CCI13 0.09* -0.15*** 0.09* -0.18*** 0.07 -0.16*** 0.09* -0.16*** -0.01 -0.17*** -0.06 0.62*** 1 NO_SERV14 0.11** 0.07 0.08 0.05 0.08 0.05 0.10* 0.06 -0.12** -0.07 0.00 -0.20*** 0.58*** 1 PREG115 -0.06 -0.04 -0.02 -0.02 -0.03 -0.05 -0.04 -0.03 0.15*** 0.09* -0.03 0.19*** -0.48*** -0.81*** 1 PREG216 -0.13* -0.10 -0.14* -0.10 -0.13* -0.07 -0.13* -0.09 -0.04 -0.03 0.05 0.10 -0.37*** -0.70*** . 1 PREG1_217 -0.12** -0.09* -0.10* -0.08* -0.11** -0.09* -0.10* -0.08 0.06 0.03 -0.00 0.18*** -0.48*** -0.82*** 0.63*** 0.98*** 1 SERV_3wk18 -0.03 -0.18*** -0.00 -0.17*** -0.03 -0.19*** -0.01 -0.18*** 0.11** 0.02 -0.08* 0.06 0.10* 0.06 0.04 0.00 0.03 1 PREG19 -0.13** -0.12** -0.10* -0.11** -0.10* -0.12** -0.12** -0.12** 0.11** 0.03 -0.01 0.06 -0.19*** -0.46*** 0.44*** 0.41*** 0.52*** 0.18*** 1 *P<0.05; **P<0.01; ***P<0.001.
1MY = total lactation milk yield; 2MY_60 = 60 days postpartum milk yield; 3PY = total lactation protein yield; 4PY_60 = 60 days postpartum protein yield; 5FY = total lactation fat yield; 6FY_60 = 60 days postpartum fat yield; 7
LY = total lactation lactose yield; 8LY_60 = 60 days postpartum lactose yield; 9P% = whole lactation protein percentage; 10P%_60 = 60 days postpartum protein percentage; 11RATIO_FP = total lactation fat to protein ratio;
12CSI = calving to submission interval; 13CCI = calving to conception interval; 14NO_SERV = number of services per conception; 15PREG1 = pregnant to first service; 16PREG2 = pregnant to second service; 17PREG1_1 = pregnant to first and second services; 18SERV_3wk = submission within 3 weeks; 19PREG = overall pregnancy rate.
39
Relationship between milk protein percentage and reproductive performance
Some significant correlations existed between protein percentage and fertility (Table
4.4). First 60 days protein percentage was negatively correlated with calving to submission interval (P<0.01) and calving to conception interval (P<0.001). A slightly positive correlation was evident between first 60 days protein percentage and first service pregnancy rate (P<0.05). Significant negative correlations were found between first 120 days protein percentage and calving to conception interval (P<0.05), and between number of services per conception and first 180 days protein percentage (P<0.05). Both first 120 days protein percentage and first 180 days protein percentage were positively correlated with first service pregnancy rate (P<0.05). No other correlations between first 120 days protein percentage or first 180 days protein percentage and fertility traits were significant. High overall protein percentage was associated with longer calving to submission interval (P<0.01), high first service pregnancy rate (P<0.001), high first three weeks submission rate (P<0.01) and high overall pregnancy rate (P<0.01), but was negatively correlated with number of services per conception (P<0.01). Overall, higher protein percentage was associated improved fertility performance of the cow, as shown by the higher submission rate within the first 3 weeks after the start of breeding season, fewer services per conception, and higher pregnancy rate.
40
Table 4.4. Correlation coefficients between milk protein percentage and measures of fertility.
Trait P%_60 P%_120 P%180 P% CSI CCI NO_SERV PREG1 PREG2 PREG1_2 SERV_3wk PREG
P%_601 1 P%_1202 0.96 1 P%_1803 0.87 0.98 1 P%4 0.75 0.90 0.97 1 CSI5 -0.13** -0.04 0.02 0.11** 1 CCI6 -0.17*** -0.10* -0.05 -0.01 0.62*** 1 NO_SERV7 -0.07 -0.08 -0.09* -0.12** -0.20*** 0.58*** 1 PREG18 0.09* 0.09* 0.10* 0.15*** 0.19*** -0.48*** -0.81*** 1 PREG29 -0.03 -0.04 -0.04 -0.04 0.10 -0.37*** -0.70*** . 1 PREG1_210 0.03 0.03 0.03 0.06 0.18*** -0.48*** -0.82*** 0.63*** 0.98*** 1 SERV_3wk11 0.02 0.04 0.06 0.11** 0.06 0.10* 0.06 0.04 0.00 0.03 1 PREG12 0.03 0.06 0.08* 0.11** 0.06 -0.19*** -0.46*** 0.44*** 0.41*** 0.52*** 0.18*** 1 * P<0.05; **P<0.01; ***P<0.001. 1
P%_60 = 60 days postpartum protein percentage; 2P%_120 = 120 days postpartum protein percentage; 3
P%_180 = 180 days postpartum protein percentage; 4P% = whole lactation protein percentage; 5CSI = calving to submission interval;
6
CCI = calving to conception interval; 7NO_SERV = number of services per conception; 8PREG1 = pregnant to first service; 9
PREG2 = pregnant to second service; 10PREG1_2 = pregnant to first and second services; 11SERV_3wk = submission within 3 weeks;
12
41
LOGISTIC REGRESSION ANALYSIS
There was a positive relationship between milk protein percentage and fertility performance of the cow. Cows with high milk protein percentage during the whole lactation period had a higher probability of becoming pregnant to the first service after the start of the breeding season (P<0.05) (Figure 4.2) and a higher probability of becoming pregnant during the whole breeding season (P=0.07) (Figure 4.3). The probability of 3-week submission rate also increased (P=0.09) with milk protein percentage over the whole lactation (Figure 4.4).
Figure 4.2. Logistic regression of the probability of a cow becoming pregnant during the first service after the start of the breeding season on milk protein percentage of the whole lactation. 0 0.2 0.4 0.6 0.8 1 2.5 3 3.5 4 4.5 Protein Percentage P ro b a b ilit y
42
Figure 4.3. Logistic regression of the probability of a cow becoming pregnant during the whole breeding season on milk protein percentage of the whole lactation.
Figure 4.4. Logistic regression of the probability of a cow receiving first service within the first three weeks after the start of the breeding season on milk protein percentage of the whole lactation.
0 0.2 0.4 0.6 0.8 1 2.5 3 3.5 4 4.5 Protein Percentage P ro b a b ilit y 0 0.2 0.4 0.6 0.8 1 2.5 3 3.5 4 4.5 Protein Percentage P ro b a b ilit y
43
LOGISTIC REGRESSION WITH PRINCIPAL COMPONENT ANALYSIS
The regression of the logit of the probability of the fertility performance on the